Search filter to identify reports of randomised controlled trials in CINAHL

Article type
Authors
Glanville J1, Dooley G2, Wisniewski S3, Foxlee R3, Noel-Storr A4
1York Health Economics Consortium
2Metaxis
3Cochrane Editorial Unit
4Cochrane Dementia and Cognitive Improvement Group
Abstract
Background:
The Cochrane CENTRAL database seeks to gather together reports of randomised controlled trials, to inform Cochrane Reviews. The Cochrane Editorial and Methods Department is developing methods to identify reviews from a range of databases to include in CENTRAL. Published search filters to identify reports of randomised controlled trials (RCTs) on CINAHL Plus are not recent. Reports of RCTs in CINAHL may not be being identified efficiently and therefore searchers may be missing studies.

Objectives:
We undertook this project to develop, test and validate a search filter to identify reports of RCTs, quasi-RCTs and controlled clinical trials from CINAHL Plus.

Methods:
We identified 11 sets of relevant and irrelevant records. We used nine sets to develop and test search filters iteratively. We used two sets to validate the filter performance in terms of sensitivity and precision. We also evaluated the performance of two previously published filters and the filter built into EBSCOhost. 

Results:
Following a sequence of iterative development we have developed a filter that offers sensitivity of 0.88 (95% confidence interval (CI) 0.77 to 0.95) and precision of 0.36 (95% CI 0.31 to 0.41). This is comparable to the sensitivity of the published filters, but represents a great improvement in terms of precision. 

Conclusions:
A sensitive and precise filter is available for use in identifying reports of RCTs, controlled clinical trials and quasi-RCTs from the CINAHL Plus database via EBSCOHost. The precision of the filter is such that it is likely to cut the number of results that need to be scanned by researchers to a third of those retrieved by a subject search alone. 

Patient or healthcare consumer involvement:
This research did not include patients or consumers.